Ehud Reiter defines four LLM paper types

The author outlines four distinct categories of LLM-assisted and LLM-written academic papers to bring clarity to discussions about appropriate generative AI usage in research. These range from Type A papers, where human authors use LLMs for brainstorming and writing improvement while carefully checking all outputs, to Type D papers, which are entirely generated by AI without the supposed authors understanding the work. The author argues that Type D papers violate fundamental scientific principles, while intermediate types present varying degrees of risk and questionable scientific rigor. Open article →